Before proceeding, you have to store your file.cas on the hard disk of the "local" computer from where you will launch Fluent. 
You will have to create a file containing the list of the computers which will make the calculations. Note that Fluent has to be launched from the directory where this file is stored.

First run the fluent command in Unix window, but without mentionning 2d or 3d, this will be asked later on. When the Fluent window prompts; select in the toolbar File, then Run...: 

A window like this should appear :

Here you can select the 3D calculation (1), (2D is default setting). Then click on parallel (2). After this you have to tell Fluent which computer it has to use. This is done with a file (in this presentation: hosts) where a listing of the computer is put. Be aware that the order of the computers is important to Fluent if you want to skip a computer for any reason...
The next action is to select the number of processes (computers) (4). Then click on Apply and finally Run. (5 & 6).
Once this is finished the nodes informations are displayed in the Fluent window. You can see that the computer b004-11 is the host node (from where you spawn and gather the data, and in this case 5 nodes (n0 to n4) where the calculations are done.

Don't forget now to load the case to process : File/Read/Case... where the file.cas/msh is located. If the case is not partitioned for the parallel computation (it should done when you load your case...), go to Parallel/Partition and select Auto partiton to share the job between each computer.

Then check your boundaries conditions, initialize the solver (Solve/Initialize) and start the iteration. (Solve/Iterate)
You can get information about the computation by clicking Parallel/Timer/usage : 
CPU time/iteration, total CPU time for the job & data transfer are displayed.